Software Project Status

1
Software Project Status

• Torre Wenaus
• BNL
• DOE/NSF Review of US LHC Software and Computing
• Fermilab
• Nov 29, 2001

2
Outline

• Project overview
• Scope, organization, planning, relation to Intl
  ATLAS
• Technical progress
• Schedule
• Budget and personnel
• Comments and conclusions

3
U.S. ATLAS Software Project Overview

• Major roles in key core software domains which
  leverage U.S. capability and are central to U.S.
  physics analysis
• Control framework and architecture
• Chief Architect, principal development role.
  Software agreement signed.
• Databases and data management
• Database Leader, major development roles
• Specify US roles via software agreements with
  International ATLAS
• Software support for development and analysis
• Software librarian, quality control, software
  development tools, training
• Subsystem software roles complementing hardware
  responsibilities
• Closely coupled to core development in a tight
  feedback loop
• Leadership roles commensurate with our activities
• Scope commensurate with U.S. in ATLAS 20 of
  overall effort

4
U.S. ATLAS Software Organization
5
U.S. ATLAS - ATLAS Coordination
US
International
US roles in Intl ATLAS software D. Quarrie
(LBNL), Chief Architect D. Malon (ANL), Database
Coordinator P. Nevski (BNL), Geant3 Simulation
Coordinator H. Ma (BNL), Raw Data
Coordinator C. Tull (LBNL), Eurogrid WP8
Liaison T. Wenaus (BNL), Planning Officer
6
ATLAS Subsystem/Task Matrix
7
Project Planning Status

• U.S. ATLAS WBS based on XProject essentially
  complete
• XProject extended to support International ATLAS
  integration in International ATLAS by Helge
  Meinhard complete
• Detailed U.S. WBS integrated into ATLAS PBS
• U.S. and ATLAS versions mostly coincide (wherever
  possible)
• ATLAS and U.S. ATLAS schedules fully integrated
• U.S. and ATLAS project management cooperating and
  in synch
• Common WBS and schedule sources in ATLAS CVS
• Projections from common sources differentiate
  the projects
• Projections also used for grid planning US Grid
  Computing WBS
• CMS, PPDG, EUDG also using/evaluating XProject

8
ATLAS Planning Officer

• The Planning Officer is responsible for
• proactively requesting and gathering schedule
  input from project coordinators
• assessing consistency with the rest of the
  project and completeness
• iterating as necessary to maintain a credible
  schedule
• maintaining the schedule and PBS using the agreed
  project management tools (currently XProject)
• presenting these materials in useful forms on the
  web, and reporting schedule status
• Ensuring adequate technical support for project
  management tools in use,
• Schedule should be reasonably detailed 1-2 years
  out, and should cover major milestones through
  the life of the project
• The Planning Officer participates by invitation
  in those parts of CSG meetings related to
  planning, and reports to the Computing
  Coordinator.

9
Progress Overview

• Control Framework and Architecture
• Databases
• Software Support and QA/QC
• Grid Software

10
Control Framework and Architecture

• Architecture review is concluded and dissolved.
  Athena is endorsed and in use throughout ATLAS
• StoreGate transient event model evolution and
  adoption has moved rapidly, with almost all
  reconstruction software now using it
• Data Dictionary prototype being developed to
  implement autogenerated event object descriptions
  and persistency mechanisms
• User and developer guides written
• ATLAS migration to CMT code management a large
  drain on developers generally and D.Quarrie in
  particular. Now completed.
• LBNL developer C.Day replaced by a postdoc, to
  save money
• FTE count constant at 5.5 (4.5 from project).
  Scripting development delayed

11
Databases

• D.Malon now sole ATLAS DB coordinator congrats!
  A positive step.
• Objectivity-based event store deployed to Lund
  users and will be used in DC0
• DB technology decision at end 2002, based on
  evaluations in DC1
• Event store architecture design document released
  in September
• Excellent basis to proceed. Fully consistent with
  both Objectivity and hybrid ROOT/RDBMS
  technologies which are to be evaluated in DC1
• One of its principal authors, Ed Frank (UC), is
  unfortunately leaving HEP
• Reduces ANL-area DB effort by 0.5 FTE to 3.5 (3.0
  supported by project)
• Event store work on the hybrid solution is
  beginning to ramp
• Prototype deployment in time for DC1s technology
  evaluation (spring)
• First component, ROOT persistency service,
  incorporated in latest release
• Progress depends on transferring expertise from
  STAR, which has a production hybrid event store.
  Requires that the planned BNL ramp proceeds.
• Hybrid event store now a hot topic. Will seek to
  establish a common project.
• Planning the process for evaluation and decision
  on event store is a top priority

12
Software Support, Quality Control

• New releases are available in the US 1 day after
  CERN
• Provided in AFS for use throughout the US
• US-developed nightly build facility used
  throughout ATLAS
• Full software build based on most recent tags
  email to developers
• Now an integral part of the release process
• Recently moved from BNL to CERN to better support
  the whole community
• Framework for progressively integrating more QC
  testing
• When CMT stabilizes, nightlies will be extended
  to incorporate more QC
• Code compliance, component testing, large scale
  chain testing
• Leveraging the experience of a recently hired QC
  expert
• Component testing proposal submitted to ATLAS and
  being prototyped
• Deploying pacman (Boston U) for remote software
  installation

13
Grid Software

• Major new US grid projects approved (PPDG SciDAC,
  iVDGL) and must be leveraged to contribute as
  much as possible to ATLAS while respecting the
  programs and deliverables of the grid projects
• Software development within the ATLAS complements
  of the grid projects is being managed as an
  integral part of the software effort
• Objective is to integrate grid software
  activities tightly into ongoing core software
  program, for maximal relevance and return
• Grid project programs consistent with this have
  been developed
• Grid goals, schedules integrated with ATLAS
  (particularly DC) program
• This does not mean that eg. PPDG FTEs can be
  subtracted from our project needs grid projects
  lead to scope extensions and priority
  redirections that are a challenge for us to
  accommodate as it is

14
Schedule

• Integrated (U.S. software U.S. grid ATLAS),
  comprehensive schedule
• Linked to U.S. ATLAS, ATLAS, U.S. Grid WBSs
  throughout
• Supports, but does not yet show most linkages
  between tasks/milestones
• Reasonable detail for near term sketchier beyond
  that
• Currently developing and adding detail for
  2002-2003, particularly Data Challenge related
• Need to return to and sustain a schedule that is
  detailed 2 years out
• Little schedule development in International
  ATLAS since the spring
• Schedule development (planning officer) a US
  responsibility since Nov 22
• WBS and schedule are input to the U.S. ATLAS
  project management accounting and tracking system

15
Summary Software Milestones
The Data Challenges will frame our objectives and
milestones in 02, 03
Slippage is apparent LHC startup delay is likely
to be swallowed by the lengthening schedule
16
Agency Budget Guidelines

• Original agency profile of monies in software
  FY01-06 was an impossible one for software
• Sharp, late peak like the profile of a failed
  project
• Fitting the profile makes critical mass at any US
  site impossible
• Dismissals of valuable HEP experts would have
  been necessary
• We developed a compromise profile well below
  our Jan 2000 proposal which provides the needed
  flatter profile
• The current agency profile is better but still
  too back-loaded, falling short until FY04
• Shortfall in FY03 could delay the ongoing
  transfer of hybrid event store expertise from
  STAR to ATLAS
• No funds until FY04 for dedicated CERN presence

17
SW Funding Profile Comparisons
2000 agency guideline
January 2000 PMP
Current agency guideline
Compromise profile requested last year
18
Budget ( Personnel) Priorities for FY01

• FY01 priorities suffered due to funding
  shortfalls
• Sustain LBNL and ANL efforts
• Highly experienced LBNL developer released, to be
  replaced by young programmer, to reduce cost
  while preserving FTE count
• Highly experienced U Chicago developer working
  with ANL DB team (50 level) is leaving. No
  resources at present to replace him
• Begin the delayed BNL ramp Add first sw pro
  developer
• First sw pro was added, but 1FTE of base support
  was lost. Temporarily compensated with lab
  resources.
• Establish sustained presence at CERN
• Unfunded. 1 person is at CERN via existing funds
  (LBNL relocation)

19
Personnel Priorities for FY02, FY03

• FY02, FY03 priorities are also in jeopardy
• Sustain LBNL (4.5FTE) and ANL (3FTE) support
• We hope the recent cutback will be sufficient.
• Add FY02, FY03 1FTE increments at BNL to reach
  3FTEs
• FY02 is in the budget. FY03 only partially fits
  in the profile. FY03 involves no new hire
  transfers ROOT expertise from STAR.
• Restore the .5FTE lost at UC to ANL
• No resources, based on profile
• Establish sustained presence at CERN.
• No resources, based on profile
• We rely on labs to continue base program and
  other lab support to sustain existing complement
  of developers

20
FY02 Software FTEs by Category
Total of 17.20
21
Software Activity in FY02
FTE breakdown by category and funding source
22
Software Funding in FY02
FTE breakdown by funding source and institute
23
FY02 Personnel
Includes grid, US support, management effort as
well as core effort.
Green new or newly active people since last
November
24
FY03 Software Project Costs
Priority
1
2
3
Guidance for FY03 is 2.4M
25
Funded FTEs based on profile
Reflects our priorities of achieving and
sustaining critical mass efforts in our
focus areas at the labs, complemented when
funding permits by dedicated effort at CERN. Note
that dedicated CERN effort peaks at 2 and rolls
over into a decline by FY06.
26
US Software Project Effort
FTEs in principal activities by fiscal year and
category.
27
Total US Core Software Effort
FTEs by fiscal year and WBS category. Total
effort from all sources.
28
Architect/Framework Needs Estimate
29
Database Needs Estimate
30
Planned Required Effort Levels
Needs are based on bottom-up estimate of Intl
ATLAS needs from WBS level 5. Developed by U.S.
software managers based on experience (developed
by one of us, reviewed by other two revisions
were small). Broadly consistent with
International ATLAS estimates.
Provided effort levels are the total of on- and
off-project sources. Grid SW FTEs are shown
separately in the DB section they do not
translate directly into ATLAS data mgmt FTEs.
31
Grid Comments

• We are following the funding mandates
• While the critical path task of event store
  development is resource starved 3 months before
  DC1, which is to include a full evaluation of
  implemented event store technologies to come to
  an already delayed decision,
• we have a substantial and growing grid effort 15
  months before the start of the first DC with grid
  objectives.
• Point 2 is as it should be -- except in the
  presence of point 1
• We try hard to use grid resources in a way that
  furthers our priorities
• eg. to develop metadata management of the sort
  needed for both distributed data management and a
  hybrid event store
• but we are constrained we must respect grid
  objectives and deliverables, and accept the
  substantial scope broadening required by working
  in the grid projects. Worthwhile in the long run,
  but not consistent with our needs in the short
  run when resources for the project proper are
  deficient.

32
Conclusions

• The project has matured, with successful programs
  at the three labs working closely with one
  another and with International ATLAS
• Scope of the program has been kept within the
  bounds of our plan
• The US continues to secure important positions in
  Intl ATLAS software
• US work is mainstream, accepted, and often of
  central importance
• We are seeing the benefits of our major program
  elements (DB, framework, event) being closely
  interrelated US developments leverage one
  another making the whole greater than the parts
  -- e.g. StoreGate leverages all three
• Resource shortages have made us dependent on a
  patchwork of funds supplementing project funds to
  meet our program, and this will continue until at
  least FY04 under the guideline profile. From FY04
  we can support our core program under the profile
  with 1-2 dedicated people at CERN (placing more
  at CERN will require relocating lab people)